In 1936, Studley noted that patients who had lost 20%

Transcription

1 GASTROENTEROLOGY 2001;121: AGA Technical Review on Parenteral Nutrition This literature review and the recommendations therein were prepared for the American Gastroenterological Association Clinical Practice and Practice Economics Committee. The paper was approved by the Committee on September 13, 2001, and by the AGA Governing Board on May 18, In 1936, Studley noted that patients who had lost 20% or more of their body weight before surgery for peptic ulcer disease had a significantly higher mortality than patients with the same condition but more modest weight losses. 1 Other observational studies subsequently confirmed this association between malnutrition (defined by weight loss, reduced anthropometric measurements, hypoalbuminemia, and/or loss of delayed cutaneous hypersensitivity) and an adverse clinical outcome in patients with a variety of diseases. 2 4 Even healthy individuals who are deprived of adequate nutrients for a long enough period of time will develop adverse clinical events because of malnutrition. 5 Therefore, it seems reasonable that providing nutritional support to malnourished patients would have beneficial effects. Years of research were directed at developing solutions and technology whereby total nutrient requirements could be safely administered intravenously. 6 In the 1960s, investigators at the University of Pennsylvania combined central venous access with newly developed nutrient solutions, 7 and the modern era of parenteral nutrition was born. Parenteral nutrition was widely accepted and implemented. Subsequently, the use of enteral nutrition and tube feeding also increased and became more sophisticated. Most patients received this nutritional support for days or weeks in the hospital; longer-term delivery was accomplished in home settings. Societies specializing in nutritional support were created throughout the world. The annual expenditure for nutritional support in the United States grew into billions of dollars. 8,9 The rationale for using nutritional support is often the same whether the intervention is delivered orally, enterally, or parenterally. However, the objective of this technical review is only to assess the clinical efficacy of parenteral nutrition. Conclusions made regarding the efficacy of this form of nutritional support may not necessarily apply to enteral nutritional support. Physiologic differences are present when nutrients are delivered directly into the systemic circulation rather than into the intestine. For example, parenteral nutrition completely eliminates enteral stimulatory processes and the first pass of nutrients through the liver. The best way to establish the efficacy of a clinical intervention is to demonstrate it in a randomized controlled trial (RCT). Therefore, we limited our analysis to data from RCTs that evaluated the efficacy of parenteral nutrition on one or more clinically important parameters of mortality, morbidity, duration of hospitalization, or cost. Whenever possible, we reviewed studies that compared the use of parenteral nutrition to no nutritional support. Because only limited data were available for some clinical conditions, other RCTs were also considered, including those that compared enteral with parenteral nutrition. This review was prepared using the principles of systematic reviewing. A protocol was written in which a priori decisions were made regarding the literature search strategy and the planned analyses. Wherever possible, we used meta-analysis to make an estimate of the treatment effect of parenteral nutrition. Meta-analysis is a statistical technique that combines results from multiple studies. It provides 2 potential advantages over simply looking at individual trials: (1) it allows a more precise estimate of the magnitude of a treatment effect, and (2) it can increase the power to identify a treatment effect that is not apparent in smaller individual studies. However, no 2 trials are precisely the same, and the problem of adding apples and oranges is always present. This issue (heterogeneity) is always a consideration and potential limitation of meta-analysis. Before beginning the formal review of over 100 RCTs, a cautionary statement must be made. The large majority of the RCTs failed to identify any benefit attributable to parenteral nutrition; some even found that parenteral nutrition was harmful. Virtually all of these trials excluded patients who were severely malnourished (defined by a large-percentage weight loss or a very low body mass Abbreviations used in this paper: AIDS, acquired immunodeficiency syndrome; ASCN, American Society for Clinical Nutrition; ASPEN, American Society for Parenteral and Enteral Nutrition; NIH, National Institutes of Health; RCT, randomized controlled trial by the American Gastroenterological Association /01/$35.00 doi: /gast

2 October 2001 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 971 index). We do not know if the results of these RCTs can be applied to severely malnourished patients. Materials and Methods Identification of Trials Since 1974, 1 of the authors (R.L.K.) has been collecting RCTs in the field of nutritional support. To date, over 1000 references have been identified. This collection was created by using the following strategies: 1. Manual search of Index Medicus 2. Manual search of individual journals (including abstracts from various society meetings) 3. References of identified papers 4. Personal contacts with other investigators 5. A computer search of the Cochrane Library (1999, issue 3) 6. A computer search of Embase (kindly provided by Alison Avenell) Meta-Analysis Inclusion criteria. Any RCT (including one only reported as an abstract) that met the following criteria was used in the meta-analyses: 1. The report explicitly stated that the patient groups were randomized. Quasi-randomized trials (allocation based on day of week, record number, or some other system whereby the group assignment was known by the investigator and/or subject before accession into the trial) were excluded. 2. The study compared treated patients (those receiving intravenous fluids containing a source of nitrogen [as amino acids or protein hydrolysate] and at least 10 kcal kg 1 day 1 of nonprotein calories) to control patients, who received no nutrient intake beyond that contained in ad libitum feedings and/or 5% (or in the case of neonatal trials, 10%) dextrose intravenously as maintenance fluid. 3. The study reported one or more outcomes of mortality, morbidity, duration of hospitalization, and/or cost. Eighty-two trials met these criteria and became the database for the parenteral nutrition meta-analysis Another 27 RCTs evaluated a form of intravenous nutrition in which nitrogen was infused along with an amount of calories that was inadequate to meet the daily requirement ( 10 kcal kg 1 day 1 ). 28,30,98, This form of intravenous nutrition, referred to as protein-sparing therapy, was separately evaluated in a protein-sparing therapy meta-analysis. For a variety of reasons, other RCTs were not included in these meta-analyses. These studies, as well as the reasons for exclusion, are listed in the Appendix. Outcome assessment. Each identified RCT was categorized by clinical condition and reviewed for outcomes of mortality, total complications, infectious complications, duration of hospitalization, and cost. For each clinical condition itemized in Table 1, data regarding disease-specific outcomes were also identified when available. Data from each RCT were abstracted twice, and differences were resolved by consensus. When a report included more than 1 eligible treatment arm, each treatment group was compared with the common control group and considered to be a separate trial. When data were presented as the total number of events, rather than the number of patients who had that event, it was assumed that there was 1 event per patient. (If the number of events was Table 1. Specific Clinical Conditions and Outcomes a Sought Perioperative (preoperative, postoperative, pre/postoperative) Major postoperative complications Individual postoperative complications (wound complications, intra-abdominal complications, pneumonia) Oncologic (chemotherapy, radiation therapy, bone marrow transplant) Tumor response Bone marrow toxicity Gastrointestinal toxicity Liver disease Encephalopathy Jaundice Gastrointestinal bleeding Liver transplantation Acute graft rejection Acute pancreatitis Pain resolution Pseudocyst/phlegmon/abscess formation Inflammatory bowel disease (ulcerative colitis, Crohn s disease) Duration diarrhea Remission rate Surgical requirement Low birth-weight infants Respiratory problems Time to achieve discharge weight Enterocolitis Diarrhea in infants and children Duration diarrhea Fluid/electrolyte problems Acquired immunodeficiency syndrome Quality of life Functional capacity Chronic pulmonary disease (chronic obstructive pulmonary disease, cystic fibrosis) Chronic renal failure Functional status Fluid/electrolyte complications Requirement for hospitalization Acute renal failure Recovery of renal function Fluid overload Burn injury Reoperation rates Skin grafting problems Other critical illness Duration of requirement for mechanical respiratory assistance Short bowel/inadequate bowel syndrome (home parenteral nutrition) Quality of life Functional capacity a As defined by the investigator.

3 972 AMERICAN GASTROENTEROLOGICAL ASSOCIATION GASTROENTEROLOGY Vol. 121, No. 4 greater than the number of patients, it was assumed that each patient had at least 1 such event.) When data were presented in graphic, rather than tabular, form, numerical data were estimated from the graph. Some endpoints (e.g., mortality) have low occurrence rates; differences can only be detected if large numbers of patients are available. Other outcomes (e.g., duration of hospitalization) have low rates of reporting. To detect such effects of parenteral nutrition or protein sparing therapy, global meta-analyses of all the eligible trials were performed. Meta-analysis was performed only when data were available from at least 3 trials. Thus, the calculations were performed only in the following sets of RCTs: 1. A global meta-analysis of all eligible trials (both parenteral nutrition and protein sparing) 2. Perioperative trials (both parenteral nutrition and protein-sparing therapy meta-analyses) 3. Oncologic therapy trials (cancer chemotherapy, radiation therapy, and bone marrow transplantation; only parenteral nutrition meta-analysis) 4. Alcoholic hepatitis trials (only protein-sparing therapy meta-analysis) 5. Low birth-weight infants trials (only parenteral nutrition meta-analysis) For dichotomous variables (mortality and morbidity events), each estimated effect is presented as the absolute risk difference (the difference between the incidence in the treated group and the incidence in the control group) and 95% confidence interval. A negative risk difference indicates that there is a decreased risk, and a positive one indicates that there is an increased risk associated with the treatment. A significant effect is present whenever a 95% confidence interval does not overlap 0. The number needed to treat to prevent (or cause) one outcome event is calculated by dividing 100 by the absolute risk difference. For example, if the risk difference for a particular complication were 5%, it would be necessary to treat 20 patients to prevent one such event. Similarly, if the risk difference were 5%, treating 20 patients would result in one additional complication. Duration of hospitalization is a continuous variable. These data are reported as weighted mean differences (a weighted average of the differences seen in each trial) and 95% confidence interval. A negative number represents a shorter duration in the treatment group. In these analyses, a significant effect is present when the 95% confidence interval does not overlap 0. Because heterogeneity was anticipated, all of the metaanalyses were performed using a random effects model. The computer programs used were Revman and Metaview 3.1 (Cochrane Collaboration, Oxford, England). Subgroup analyses. For each of the outcomes in the parenteral nutrition meta-analysis, separate calculations were undertaken to assess the following factors: 1. Use or non-use of lipid in the parenteral nutrition formulation. Nutritional support was considered to include lipid if at least 15% of total nonprotein calories in the nutritional formulation were derived from lipid. Nonuse of lipid meant either that no lipid was provided or that only small amounts of lipid ( 15% of the total nonprotein calories) were used periodically, usually to prevent essential fatty acid deficiency. 2. Presence or absence of malnutrition in the study population. A trial was considered to contain malnourished patients if at least 50% of the patients satisfied the investigators definition of malnutrition at entry. A trial was considered not to contain malnourished patients if 50% of the patients satisfied this criterion. Malnutrition was variably defined in these trials, and those definitions ranged from modest weight loss alone to more profound weight loss, and/or hypoalbuminemia, and/or abnormalities in anthropometric measurements, or skin test reactivity. 3. Inclusion of only those trials that were reported in full paper form (i.e., not just as an abstract). 4. Inclusion of only those trials in which the nutritional therapy was provided for at least 7 days. 5. Inclusion of only those trials that reported outcome events as the number of affected patients (i.e., not just as total number of events). 6. Provision of parenteral nutrition only in the preoperative period, only in the postoperative period, or in both. 7. Provision of parenteral nutrition only for patients undergoing surgery for upper gastrointestinal cancer. (This subgroup analysis was the only one not determined a priori.) 8. Provision of parenteral nutrition for patients treated with chemotherapy, radiation therapy, or bone marrow transplantation. We were unable to perform such subgroup analyses in the meta-analysis of protein-sparing therapy for the following reasons: 1. The caloric content was not an issue. 2. Preoperative or pretreatment nutritional status was not usually reported or was normal. 3. All but 1 RCT was published as a full paper. 4. Almost all of the surgical trials provided postoperative protein-sparing therapy for 7 days, and all of the trials in alcoholic hepatitis used it for more than 1 week. Results Global Meta-analysis of Parenteral Nutrition Trials. A total of 82 RCTs was identified The calculated absolute risk differences (or weighted mean average) and the respective confidence intervals for the principal mortality and morbidity parameters, as well as duration of hospitalization, are summarized in Table 2. The

4 October 2001 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 973 Table 2. Global Meta-Analyses Outcome Absolute risk difference a Confidence intervals Number of studies (patients) included Parenteral nutrition Mortality 0% 2%, 1% 69 (3750) Total complication rate 6% b 4%, 16% 47 (2685) Infectious complication rate 5% c 1%, 9% 57 (2832) Duration hospitalization 0.10 days d 1.83 days, 1.64 days 13 e (504) Protein-sparing therapy Mortality 0% 2%, 3% 18 (1218) Total complication rate 3% 7%, 1% 19 (1271) Infectious complication rate 2% 1%, 5% 18 (1178) Duration hospitalization 0.2 days 0.72 days, 1.04 days 10 f (246) a This represents the difference between the outcome in the treated group and the control group; a negative number represents a benefit for the treated group. b In trials in which lipid was not used, absolute risk difference equaled 26% ( 4%, 48%). c Data available to do separate meta-analyses in perioperative patients, oncologic therapy patients, and low birth-weight infants. See Tables 3, 5, and 9. Six other trials provided data regarding infection rates in patients with alcoholic hepatitis, 96 pancreatitis, 99 colitis, 100 AIDS, 107 cystic fibrosis, 109 and trauma 111 ; when these combined, absolute risk difference equaled 10% ( 2%, 17%). d Subgroup analyses identified the following significant absolute risk differences: Use of lipid: 1.7 days ( 3.1 days, 0.3 days); nonuse of lipid: 5.3 days ( 1.4 days, 9.1 days); including malnourished patients: 5.0 days ( 1.6 days, 8.4 days); not including malnourished patients: 1.8 days ( 3.2 days, 0.3 days). e 12 of these trials in perioperative patients. f All 10 trials in surgical patients. footnotes in this and subsequent tables include data from selected subgroup analyses. Mortality. Parenteral nutrition did not influence mortality. The subgroup analyses did not alter the overall estimate. Complications. There was no significant effect of parenteral nutrition on the total complication rate. Only 1 of the subgroup analyses identified a significant difference in the trials in which lipid was not used, parenteral nutrition was associated with an increased total complication rate. Parenteral nutrition was associated with a significant increase in the infectious complication rate; the absolute risk difference was 5%. Parenteral nutrition resulted in 1 additional infection for every 20 patients who were treated. In almost every subgroup analysis, the estimates were positive (treatment associated with more infections), although the confidence intervals sometimes overlapped 0. Significant differences were observed in the trials in which lipid was not used, in which only nourished patients were included, in which parenteral nutrition was provided for at least 7 days, and in the trials reported as full papers. This harmful effect was caused largely by the effect parenteral nutrition had in cancer patients receiving oncologic therapy. It was not observed when the perioperative or low birth-weight clinical conditions were separately analyzed. Six other trials of parenteral nutrition provided data regarding infection rates. 96,99,100,107,109,111 When these RCTs were combined, a treatment-related increase in the infection rate was also observed. This latter computation is displayed in Figure 1. Figure 1. Rates of infectious complications in 6 trials not included in meta-analyses of perioperative, oncologic, or low birth-weight infants. In all 6 trials, the direction of the effect is positive (control group had better outcome). Abbreviations: Expt, treated group; Ctrl, control group; RD, risk difference; n, number with infection; N, number randomized to that arm; CI, confidence interval; df, degrees of freedom; Z, Z score for heterogeneity.

5 974 AMERICAN GASTROENTEROLOGICAL ASSOCIATION GASTROENTEROLOGY Vol. 121, No. 4 Duration of hospitalization. Parenteral nutrition had no major effect on duration of hospitalization, at least in the perioperative state. Using lipid in the nutritional infusion and only treating groups of nourished patients were the 2 factors significantly associated with reduced hospital stays of about 2 days. Nutritional infusions not including lipid and treatment of populations that included malnourished patients were the 2 factors significantly associated with prolonged durations of stay of about 5 days. Cost. There were inadequate data to perform a meta-analysis regarding cost. Summary and implications of the data. The problem of heterogeneity is most apparent in this global meta-analysis. At first consideration, it would not appear to make any sense to combine trials in so many disparate clinical conditions, even if the therapeutic intervention being tested was similar. We decided a priori to undertake this meta-analysis despite the heterogeneity of the trials because we believed that this would be the only way to identify a subtle effect of parenteral nutrition (e.g., on mortality). It was our intent to do a series of subgroup analyses to identify more precisely the circumstances under which effects were or were not present. We subjected 82 RCTs (containing over 4000 patients) to meta-analysis. The results indicated that, compared with standard therapy, parenteral nutrition did not improve survival, the total complication rate, or the length of hospitalization. The meta-analysis showed that parenteral nutrition increased the infectious complication rates. In part, this was attributable to an adverse effect in cancer patients undergoing oncologic therapy. Such an adverse effect was not observed in perioperative patients or low birthweight infants. However, the fact that this effect was also present when the other trials were combined indicates that the relationship between parenteral nutrition and an increased risk of infection is more generalizable. Global Meta-analysis of Protein-Sparing Therapy Trials. There were 27 trials that compared protein sparing to standard therapy. 28,30,98, The results of the meta-analyses of this intervention are presented in Table 2. Mortality. Protein-sparing therapy did not affect survival. Complications. Protein-sparing therapy had no effect on the total or infectious complication rates. Duration of hospitalization. Data regarding the duration of hospitalization were available only in surgical trials. Protein-sparing therapy had no effect on this outcome. Cost. There were inadequate data to perform a meta-analysis regarding cost. Summary and implications of the data. This meta-analysis of 27 RCTs did not show any beneficial effect of protein-sparing therapy on mortality, morbidity, or duration of hospitalization. Most of these trials were conducted in perioperative patients or in patients with alcoholic hepatitis. Perioperative Trials Trials. The largest number of RCTs in both meta-analyses came from trials of patients treated before and/or after surgery. There were 41 such trials identified for the parenteral nutrition meta-analysis, and 20 RCTs evaluated protein-sparing therapy. 28,30, The largest RCT identified in this review was one of the protein-sparing therapy studies; it included 678 patients. 114 The estimated absolute risk differences for both parenteral nutrition and protein-sparing therapy are presented in Table 3. Mortality. Neither parenteral nutrition nor protein-sparing therapy affected postoperative mortality. The subgroup analyses did not identify any significant beneficial or harmful effects. Postoperative complications. Parenteral nutrition had no significant effect on the various postoperative complication rates. However, the directions of the absolute risk differences were always negative (i.e., the effect tended to favor treatment). A number of subgroup analyses were performed. Parenteral nutrition was associated with a significant reduction in total complications in the trials that excluded malnourished patients and in trials in which lipids were part of the nutritional formulation. When the trials were separated into those that provided only preoperative, only postoperative, or both preoperative and postoperative parenteral nutrition, no significant effect was observed, although the absolute risk differences tended to be slightly more favorable for treatment in the preoperative trials. When only trials containing patients with upper gastrointestinal cancer were considered, the absolute risk differences were always negative. In this subgroup, parenteral nutrition significantly reduced the major complication rate. Postoperative patients are given nothing by mouth for some period of time. To see if longer periods of therapy were important, we separately analyzed the subset of 13 postoperative trials in which the treatment groups were administered parenteral nutrition for more than 7 days

6 October 2001 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 975 Table 3. Meta-Analyses of Perioperative Trials Outcome Absolute risk difference a Confidence intervals Number of studies (patients) included Parenteral nutrition Mortality 0% 2%, 2% 37 (2164) Total complications 6% b,c 13%, 1% 32 (2062) Infectious complications 2% 8%, 3% 29 (1612) Major complications 3% d 9%, 3% 22 (1648) Wound complications 2% 6%, 2% 29 (1800) Intra-abdominal complications 0% e 5%, 4% 21 (1375) Postoperative pneumonia 2% 4%, 2% 23 (1684) Protein-sparing therapy Mortality 1% 2%, 3% 13 (1033) Total complication rate 3% 7%, 1% 16 (1182) Infectious complication rate 2% 1%, 5% 16 (1109) Major complications 1% 6%, 4% 6 (773) Wound complications 0% 4%, 3% 11 (1005) Intra-abdominal complications 1% 3%, 5% 10 (950) Postoperative pneumonia 1% 2%, 3% 10 (957) a This represents the difference between the outcome in the treated group and the control group; a negative number represents a benefit for the treated group. b In trials that excluded malnourished patients, absolute risk difference equaled 18% ( 34%, 2%). c In trials in which lipid was used, absolute risk difference equaled 7% ( 15%, 0.2%). d In trials which only included patients with upper gastrointestinal cancers who were undergoing surgery, absolute risk difference equaled 18% ( 31%, 6%). e Actual estimate is 0.3%, therefore estimated risk differences for each of the complication rates in the parenteral nutrition meta-analysis were negative. (range, 8 18 days) and the control group was not given any form of nutritional support. 22,24,29,33,37 40,47 One of these trials did allow fallback to parenteral nutrition, 24 but only after a complication occurred. The results of the meta-analysis of these trials are displayed in Table 4. No significant effect of the nutritional intervention was observed. Protein-sparing therapy had no effect on postoperative morbidity. The results from the large single trial 114 were the same as the estimates derived when all of the smaller RCTs were combined. Duration of hospitalization. Parenteral nutrition had no effect on the duration of hospitalization of surgical patients. Summary and implications of the data. Although parenteral nutrition was not shown to reduce postoperative complications with statistical certainty, almost all of the absolute risk differences were negative. Even if a true beneficial effect were to exist, it would not be dramatic. A small benefit must be weighed against the cost of the therapy. If we were to assume that parenteral nutrition reduces the total complication rate by 6% (Table 3), more than 16 patients would need to be treated for at least 7 days each to prevent 1 adverse event. The use of parenteral nutrition would be economically justifiable only if these complications were very expensive. 131 In a meta-analysis evaluating the clinical efficacy of parenteral nutrition in critically ill and perioperative patients, Heyland et al. 132 concluded that studies using lipid-free infusions and those including malnourished patients (as defined by each investigator or, if not de- Table 4. Meta-Analysis of Postoperative Trials in Which Patients Randomized to Parenteral Nutrition or No Nutritional Support for Periods of Time in Excess of 1 Week a Outcome Absolute risk difference b Confidence intervals Number of studies (patients) included Mortality 2% 1%, 6% 12 (482) Total complication rate 4% 13%, 22% 11 (477) Infectious complication rate 1% 11%, 10% 11 (477) Major complications 0% 21%, 20% 8 (376) Wound complications 2% 2%, 5% 9 (411) Intra-abdominal complications 3% 8%, 15% 9 (416) Postoperative pneumonia 3% 9%, 2% 8 (395) a References 22, 24, 29, 33, 37 40, 47. b This represents the difference between the outcome in the treated group and the control group; a negative number represents a benefit for the treated group.

7 976 AMERICAN GASTROENTEROLOGICAL ASSOCIATION GASTROENTEROLOGY Vol. 121, No. 4 fined, by 10% weight loss) were more likely to show a benefit from parenteral nutrition. We could not confirm either of these observations. In fact, in our analysis perioperative patients who received lipids had better outcomes, and trials that evaluated only well-nourished patients were more likely to demonstrate a benefit of parenteral nutrition. The probable reason for this disagreement is that the 2 meta-analyses did not combine the same data. Heyland et al. included 26 RCTs. However, 3 of them were not in surgical patients. 99,116,133 Six of the remaining 23 were excluded from this analysis , Although the remaining 17 were included in our meta-analysis, we also included 24 others. 11,12,14,17 19,23,25,26,28,29,31 38,40,42,44 46 The VA cooperative study suggested that severely malnourished patients (as defined by the Nutrition Risk Index 48 or the Subjective Global Assessment 4 ) may have benefitted from preoperative parenteral nutrition. 48 Although the differences did not achieve statistical significance, the rates of major postoperative complications were 20% 25% in severely malnourished patients given parenteral nutrition and 40% 50% in the nutritionally comparable controls. It is important to realize that this degree of malnutrition was not present in most or all of the patients enrolled in other trials. As such, our data are inadequate to confirm or refute this conclusion from the VA trial. Muller et al. 17 suggested that parenteral nutrition was beneficial in patients with upper gastrointestinal malignancies. We combined the trials that only included patients with cancer of the stomach and/or esophagus. 13,17,25,26,37 39 For each of the analyses, the estimated absolute risk differences were negative (i.e., favoring treatment), and a significant benefit (estimated risk difference, 18%) was demonstrated for reduction of major complications. Treating 5.5 patients with parenteral nutrition for at least 7 days to prevent a major complication may be justifiable from the perspective of resource allocation. However, it should be remembered that this subgroup analysis was undertaken to validate the observation of Muller et al. When the data of Muller et al. were removed, no significant differences remained, although all of the risk differences still favored treatment. It is unlikely that inadequate duration of treatment was responsible for the failure to demonstrate a more profound effect of parenteral nutrition. Most of the trials provided parenteral nutrition for at least a week; and when only those studies were considered, no benefit was observed. Even in the RCTs in which parenteral nutrition was provided postoperatively for more than 7 days, no effect of the intervention could be demonstrated. A review of nutritional support, sponsored by the National Institutes of Health (NIH), the American Society for Parenteral and Enteral Nutrition (ASPEN), and the American Society for Clinical Nutrition (ASCN) was published in and included an evaluation of the influence of parenteral nutrition on clinical outcome in surgical patients. Pooling data separately from 13 RCTs that provided preoperative parenteral nutrition and 8 RCTs that provided it only postoperatively found that the preoperative intervention resulted in a 10% reduction in postoperative complications, but that postoperative treatment alone produced a 10% increase in the complication rate. We examined this observation by considering separately the trials that used parenteral nutrition only in the preoperative period, only in the postoperative period, or during both periods. We did not find that preoperative therapy conferred significant benefits or that postoperative treatment was harmful. However, the absolute risk differences were somewhat more favorable in the preoperative studies. Again, the likely reason for the disparate findings relates to the trials that were included or excluded in the 2 analyses. We reclassified 41,43 48 or excluded 134,136, preoperative and 6 postoperative RCTs included in the NIH/ASPEN/ASCN review. In addition, we included 3 other preoperative 11,12,14,19 and 20 other postoperative 23,25 39 trials in our analysis that were not included in the NIH/ASPEN/ASCN report. Of particular note, we excluded the largest published postoperative trial (300 patients) because the control group received 10% dextrose. 136 That study found a statistically significant increase in complications in the recipients of the parenteral nutrition and was a dominant factor in the NIH/ASPEN/ ASCN analysis. Although protein-sparing therapy improves nitrogen balance, 28,30, ,118 20,124 it does not affect mortality or morbidity when it is provided during the postoperative period. This dissociation between nutritional and clinical outcomes, observed in other trials of nutritional support, 142 underscores the importance of using clinical outcomes data rather than surrogate nutritional markers. Oncologic Therapy Trials. We identified 19 RCTs assessing parenteral nutrition in cancer patients receiving chemotherapy, RCTs in cancer patients being treated with radiation therapy with or without concomitant chemotherapy, and 4 RCTs in patients undergoing bone marrow transplantation The results of this meta-analysis are summarized in Table 5.

8 October 2001 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 977 Table 5. Meta-Analysis of Oncologic Trials Outcome Absolute risk difference a Confidence intervals Number of studies (patients) included Mortality b 0% 5%, 5% 19 (1050) Total complication rate 40% 14%, 66% 8 (333) Infectious complication rate 16% 8%, 23% 18 (823) Tumor response 7% c 12%, 1% 15 d (910) Bone marrow toxicity 22% 10%, 54% 3 (134) Gastrointestinal toxicity 1% 9%, 11% 6 (310) a This represents the difference between the outcome in the treated group and the control group; a negative number represents a benefit for the treated group. b Although 1 bone marrow transplantation trial reported an improved survival, 94 this was not demonstrated when all 4 trials were combined; absolute risk difference equaled 5% ( 14%, 5%). Only 3 of these trials provided parenteral nutrition during the time when the transplantation was performed ; when only these 3 trials were combined, absolute risk difference equaled 9% ( 22%, 4%). c A negative absolute risk difference indicates that the response rate in the control group was higher than in the recipients of the parenteral nutrition. d 13 of these 15 RCTs were chemotherapy trials. Mortality. There was no apparent effect of parenteral nutrition on mortality. When subgroup analyses were performed, none could be found in which there was an effect on survival. Although Weisdorf et al. 94 reported that parenteral nutrition increased survival in patients undergoing bone marrow transplantation, we did not observe a significant survival difference when we combined the data from 4 RCTs assessing bone marrow transplantation However, the largest of the 4 RCTs 91 assessed the role of home parenteral nutrition after the patients were discharged. This is a different scenario than the one that was tested by Weisdorf et al. When the data from the other 3 were combined, there was a nonsignificant trend for the parenteral nutrition to improve mortality. One RCT did assess the value of parenteral nutrition provided to patients with end-stage malignancies receiving no specific cancer therapy. 81,82 The treated recipients lived significantly longer (46 days) than did the controls (7 days). Complications. The use of parenteral nutrition resulted in an increased total complication rate. Parenteral nutrition also increased the incidence of infectious complications. A significant increased risk for both total and infectious complications was associated with parenteral nutrition in almost every subgroup evaluated in the subgroup analyses. Tumor-response rate. The tumor-response rates (to chemotherapy in particular) were also adversely affected by the parenteral nutrition. All of the absolute risk differences in the subgroup analyses favored the control group, although these differences did not always achieve statistical significance. Treatment toxicity. Parenteral nutrition did not appear to provide any benefit with regard to either bone marrow or gastrointestinal toxicity. No effect was observed in any of the subgroup analyses. Summary and implications of the data. Parenteral nutrition does not alter survival in patients receiving radiation or chemotherapy. The data cannot exclude the possibility that in-hospital parenteral nutrition will favorably affect survival in patients undergoing bone marrow transplantation. In all other aspects, the use of parenteral nutrition in cancer patients receiving chemotherapy, radiation therapy, or bone marrow transplantation was clearly associated with net harm. Parenteral nutrition was associated with increases in total and infectious complication rates. In addition, parenteral nutrition was associated with an impaired tumor response to chemotherapy, which may be related to exogenous nutrients stimulating tumor growth. 14, Liver Disease Trials. Two trials comparing parenteral nutrition 96,97 and 5 trials comparing protein-sparing therapy to standard treatment were conducted in patients with alcoholic hepatitis. These data are summarized in Table 6. Two other RCTs evaluated patients undergoing liver transplantation. One included 2 treatment arms (different intravenous regimens) and a control group receiving no nutritional support. 25 The other compared parenteral to enteral nutrition. 147 Several trials compared parenteral nutrition that was enriched with branched-chain amino acids to a variety of other interventions in patients with hepatic encephalopathy Mortality. Neither parenteral nutrition nor protein-sparing therapy improved survival. Morbidity. With regard to alcoholic hepatitis, the only data regarding total complications came from 2 of the protein-sparing therapy trials 126,129 ; no effect was demonstrated. There were nonsignificant trends in both

9 978 AMERICAN GASTROENTEROLOGICAL ASSOCIATION GASTROENTEROLOGY Vol. 121, No. 4 Table 6. Parenteral Nutrition and Protein-Sparing Therapy in Alcoholic Hepatitis a Outcome Simon 97 Naveau 96 Protein-sparing therapy Mortality 27%/18% 5%/5% 6% ( 18%, 5%) b Total complication rate No data No data 5%/0%, c 6%/22% c Serum bilirubin Trend to be better in PN d Better e in PN group Better e in PST group in 3 of 5 trials Encephalopathy developed or worsened 0%/11% 10%/20% Assessment not possible from available data PN, parenteral nutrition; PST, protein-sparing therapy. a Numerator, rate of outcome in treated group; denominator, rate of outcome in control group. b Estimated risk difference and confidence interval (negative number represents a benefit for the treated group) from a meta-analysis of 4 trials 125,126,128,129 encompassing 155 patients. c Data from only 2 trials containing and patients, respectively. d P e P trials for parenteral nutrition to be associated with less encephalopathy. 96,97 In most of these RCTs of alcoholic hepatitis, hyperbilirubinemia improved more rapidly in those who received either form of intravenous nutrition. Parenteral nutrition had no demonstrable impact on the course of patients receiving liver transplants. In 1 of the trials, 25 the control patients were in the intensive care unit for a longer time (6 days) than were the patients in the other 2 groups (3 4 days). No data regarding graft rejection were provided. Hepatic encephalopathy improved more often and/or more rapidly in the recipients of the branched-chain amino acid enriched solutions. No data were available regarding what effect, if any, parenteral nutrition or protein sparing has on gastrointestinal bleeding in patients with liver disease. Summary and implications of the data. Only a limited number of trials are available. There is no apparent effect of parenteral nutrition or protein-sparing therapy on mortality. Parenteral nutrition may result in a more rapid decline in hyperbilirubinemia, but the clinical relevance of such an improvement is undefined. The role of parenteral nutrition after liver transplantation is unknown. The trials were so small that marked differences between groups would be needed to achieve statistical significance. A previously published meta-analysis concluded that intravenous branched-chain amino acid enriched solutions (alone or as part of a parenteral nutrition regimen) are of benefit in treating hepatic encephalopathy. 156 However, the expense of the nutritional intervention is higher than that of other available nonnutritional ones. 131 Acute Pancreatitis Trials. Two RCTs evaluated the role of parenteral nutrition in patients with acute pancreatitis. One of them 98 included a second treatment arm (protein-sparing therapy). Two other small trials compared parenteral to enteral nutrition. 157,158 All of these trials excluded patients with severe disease. These data are summarized in Table 7. Mortality. No differences in mortality were observed between the patients who received parenteral nutrition and the controls. Morbidity. Compared with standard therapy, parenteral nutrition resulted in more complications and a longer hospitalization in 1 trial. 99 Compared with enteral nutrition, parenteral nutrition was associated with a higher complication rate in 1 trial. 158 Parenteral nutrition was not shown to have any effect on the development of pseudocysts, abscesses, phlegmons, or pain resolution. Duration of hospitalization and costs. The patients receiving parenteral nutrition had significantly longer hospitalizations and higher costs in 1 RCT. 99 Summary and implications of the data. Only a few RCTs have evaluated the clinical efficacy of parenteral nutrition in patients with acute pancreatitis. Two studies compared parenteral nutrition to standard therapy. One found no statistically significant effect from parenteral nutrition or protein-sparing therapy, and the other found that the use of parenteral nutrition prolonged the length of hospitalization, increased cost, and, possibly caused more complications. Two small RCTs compared enteral with parenteral nutrition. One found no difference, and the other noted an increased rate of complications in the recipients of parenteral nutrition. It is difficult to draw firm conclusions when the only data available are from a few small trials. However, the data suggest that parenteral nutrition not only is not beneficial, but also causes net harm in patients with mild pancreatitis. We are not aware of any RCTs that evaluated the efficacy of parenteral nutrition in patients with more severe disease.

10 October 2001 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 979 Table 7. Intravenous Nutrition and Acute Pancreatitis Study (reference) Mortality Total complications Infectious complications Duration of hospitalization Floch 98 PN 0/7 (0%) PST 0/8 (0%) Control 2/8 (25%) Sax 99 PN 1/29 (3%) 6/29 (21%) a 3/29 (10%) 16 days b Control 1/26 (4%) 0/26 (0%) a 0/26 (0%) 10 days b McClave 157 PN 0/16 (0%) 2/16 (13%) 10 days EN 0/16 (0%) 2/16 (13%) 12 days Kalfarentzos 158,c PN 2/20 (10%) 15/20 (75%) b 10/20 (50%) b 39 days EN 1/18 (6%) 8/18 (44%) b 5/18 (28%) b 40 days c PN, parenteral nutrition; PST, protein-sparing therapy; EN, enteral nutrition. a P 0.05 (Assuming the 6 complications were each in different patients). b P c Although investigators described the degree of pancreatitis as severe, this determination was based on APACHE scores ( 8), levels of C-reactive protein, and CT appearances. None of these criteria are well-validated predictors of outcome in pancreatitis. Inflammatory Bowel Disease Trials. Two RCTs compared parenteral nutrition to no nutritional therapy in patients with acute ulcerative or Crohn s colitis. 100,101 Several RCTs compared parenteral to enteral nutrition in patients with acute ulcerative colitis 159 or during exacerbations of Crohn s disease The data are summarized in Table 8. Mortality. There was no impact of parenteral nutrition on mortality when compared with no nutritional support. 100,101 No data were available from the other trials. Morbidity. No effect on disease remission or subsequent need for surgery was observed when parenteral nutrition was compared with no nutritional support or to enteral nutrition. None of these trials addressed the duration of diarrhea. Summary and implications of the data. Parenteral nutrition provided no benefit in the treatment of Crohn s or ulcerative colitis, and it may be equivalent to enteral nutrition with regard to treating patients with active Crohn s disease of the small intestine. Parenteral nutrition could thus be inferred to be inferior to standard therapy, as 3 separate meta-analyses have concluded that clinical remission is achieved more frequently with corticosteroids than with enteral nutrition One additional trial evaluated the role of bowel rest; both treatment groups received parenteral nutrition and steroids, but 1 of the groups also was allowed to eat. 166 No differences were observed. All of these data indicate that bowel rest is not necessary to achieve remission in Crohn s disease. Pediatrics Trials. The RCTs of parenteral nutrition that were identified in the pediatric literature included conditions that have already been discussed perioperative patients 32 and patients receiving oncologic therapy. 60,85 88 Table 8. Parenteral Nutrition and Inflammatory Bowel Disease Mortality Remission Surgery required Study PN Control PN Control PN Control Control group received no nutritional support Dickinson 100,a 0/19 (0%) 1/17 (6%) 10/19 (53%) 11/17 (65%) 9/19 (47%) 6/17 (35%) McIntyre 101,a 1/27 (4%) 1/20 (5%) 16/27 (59%) 12/20 (60%) 11/27 (41%) 5/20 (25%) Control group received enteral nutrition Gonzalez-Huix 159,b 10/21 (48%) 12/23 (52%) 11/21 (52%) 11/23 (48%) Greenberg 160,c 12/21 (57%) 11/19 (58%) 9/21 (43%) 2/19 (11%) Alun Jones 161,c 14/19 (74%) 11/17 (65%) 1/19 (5%) 1/17 (6%) Wright 162,c 4/5 (80%) 3/6 (50%) PN, parenteral nutrition. a Colitis (ulcerative or Crohn s). b Ulcerative colitis. c Crohn s disease.

11 980 AMERICAN GASTROENTEROLOGICAL ASSOCIATION GASTROENTEROLOGY Vol. 121, No. 4 Five RCTs evaluating the effect of parenteral nutrition on low birth-weight infants were subjected to meta-analysis; the estimated absolute risk differences for mortality and morbidity are presented in Table 9. One other RCT assessed protein-sparing therapy in low-birthweight infants, 130 and 1 trial compared different levels of caloric delivery. 167 Mortality. Parenteral nutrition did not have a significant impact on survival. In the trial of protein-sparing therapy, which included only 30 patients, the mortality rates tended to be higher in the treatment group (47%) than in the control group (27%), but the difference was not statistically significant. Complications. Parenteral nutrition did not affect either the total or infectious complication rates. No effect of parenteral nutrition was observed with regard to the development of enterocolitis 104,106 or respiratory problems. 104 Infectious complications developed in 40% of the patients receiving protein-sparing therapy and 27% of the controls. Although the survivors in the proteinsparing therapy group had significantly fewer apneic episodes than did the controls who lived, most of those who died required intubation and respiratory support, and there were more deaths in the treated group. Duration of hospitalization. Additional nutrients should promote more rapid weight gain, and a subsequent earlier discharge, in neonates who are hospitalized only because of low birth weight. However, more rapid weight gain did not translate into an earlier discharge. 103,167 Summary and implications of the data. No benefits from parenteral nutrition were observed. Although this nutritional intervention may increase the rate of weight gain, there is no evidence that that translated into any meaningful clinical outcome, including an earlier hospital discharge. It is generally believed that such neonates require early nutritional support because of (1) a lack of intrinsic energy stores and (2) nutrient requirements for growth and development. For many of these infants, enteral nutrition may suffice. We have no direct data assessing how long such babies can endure starvation. Acquired Immunodeficiency Syndrome Trials. One trial randomized 31 malnourished patients with acquired immunodeficiency syndrome (AIDS) into a group receiving home parenteral nutrition for 2 months or into a group receiving nutritional counseling. 107,108 Mortality. No difference in survival was reported in the initial report; 3 patients in each group died. 107 Subsequently, the investigators reported that survival was improved in the recipients of parenteral nutrition, 108 but these data were not analyzed in accordance with the initial treatment assignment. Morbidity. The recipients of the parenteral nutrition had improved anthropometric measurements and weight gain, and they subjectively felt better. No difference in the incidence of AIDS-related complications was observed. Summary and implications of the data. This single RCT did not show that parenteral nutrition altered the progression of AIDS. It did improve body weight and subjective feelings of well-being, but the study was not blinded. Pulmonary Disease Trials. Two trials assessed the role of parenteral nutrition in pulmonary disease. One assessed the role of home parenteral nutrition in patients with cystic fibrosis. 109 The other assessed malnourished hospitalized patients with chronic obstructive pulmonary disease. 110 The results from these 2 RCTs are summarized in Table 10. Mortality. No significant differences were observed, although the mortality rates in the control groups of both trials were higher than those in the recipients of parenteral nutrition. Morbidity. The rates of catheter infections were not significantly different. 109 Table 9. Meta-Analysis of Trials of Parenteral Nutrition in Low Birth-Weight Infants Outcome Absolute risk difference a Confidence intervals Number of studies (patients) included Mortality 4% 8%, 17% 4 (207) Total complication rate b,c 42% 22%, 105% 3 (153) Infectious complication rate 1% 8%, 9% 4 (174) a This represents the difference between the outcome in the treated group and the control group; a negative number represents a benefit for the treated group. b Two RCTs provided the rates of enterocolitis in the treated/control groups; these were 5%/5% 104 and 5%/0% 106 respectively. c One report indicated that respiratory problems developed in 10% of the treated group and 0% of the controls.